Hayhooks MCP Server

What is Hayhooks?

Hayhooks makes it easy to deploy and serve Haystack Pipelines and Agents as REST APIs, and expose them over the MCP (Model Context Protocol), A2A protocol, and OpenAI-compatible endpoints. It is designed for developers who want to integrate Haystack pipelines into AI development environments like Cursor or Claude Desktop with minimal boilerplate.

How to use Hayhooks?

Install with pip install hayhooks, then run hayhooks run. Create a pipeline wrapper class inheriting from BasePipelineWrapper, deploy it with hayhooks pipeline deploy-files -n <name> <directory>, and call the resulting HTTP endpoints or interact via MCP/A2A clients. Optionally add Chainlit UI with hayhooks run --with-chainlit or tracing with pip install "hayhooks[tracing]".

Key features of Hayhooks

• Easy deployment of Haystack pipelines and agents as REST APIs
• Expose pipelines as MCP tools for AI development environments
• Support for A2A protocol for inter-agent discovery and task delegation
• Embedded Chainlit chat UI with streaming and pipeline selection
• OpenAI-compatible chat completion endpoints with streaming support
• OpenTelemetry tracing with a local dashboard for lifecycle visibility

Use cases of Hayhooks

• Deploy Haystack agents and pipelines as REST APIs for production use
• Expose Haystack pipelines as MCP tools in Cursor or Claude Desktop
• Integrate Haystack agents with Open WebUI as streaming chat backends
• Build inter-agent workflows by exposing pipelines via the A2A protocol
• Control Hayhooks core API endpoints by chatting with an MCP client